FeedbackEval introduces a systematic benchmark for measuring how well large language models understand and utilize different types of feedback when repairing code.

• Evaluates LLMs' ability to comprehend error messages, test cases, and natural language explanations
• Measures performance across diverse programming tasks and feedback types
• Provides insights into how different models handle feedback-driven code repair
• Establishes a standardized framework for consistent evaluation

This research is critical for software engineering teams seeking to integrate AI-powered code assistance tools that can effectively respond to developer feedback, potentially reducing debugging time and improving code quality.

FeedbackEval: A Benchmark for Evaluating Large Language Models in Feedback-Driven Code Repair Tasks